S A ๐Ÿ‘‘ (@samsonaligba) header
S A ๐Ÿ‘‘ (@samsonaligba) profile picture

Experienced #tech founder/operator/life long learner and force multiplier. || Always willing to contribute ๐Ÿ’ฏ

Joined April 2011
Photo by S A ๐Ÿ‘‘ (@samsonaligba)Photo by S A ๐Ÿ‘‘ (@samsonaligba)Photo by S A ๐Ÿ‘‘ (@samsonaligba)Photo by S A ๐Ÿ‘‘ (@samsonaligba)Photo by S A ๐Ÿ‘‘ (@samsonaligba)Photo by S A ๐Ÿ‘‘ (@samsonaligba)Photo by S A ๐Ÿ‘‘ (@samsonaligba)Photo by S A ๐Ÿ‘‘ (@samsonaligba)Photo by S A ๐Ÿ‘‘ (@samsonaligba)Photo by S A ๐Ÿ‘‘ (@samsonaligba)
If you create value for enough people, you will eventually become wealthy. Wealth is value creation, selah!
1
28
2
56
S A ๐Ÿ‘‘ retweeted
Meet Nora, our next AI persona. We asked Nora to handle the introduction. She had fun with it. You can chat with Nora now in our research preview.
80
59
62
510
367,512
Take this same concept of an agent harness that owns the loop and verification tool chain but for non coding tasks. That's the new frontier, domains outside of software need long running agents that self-verify. The domain experts only need to compose the verification and walk away.
Tobi Lรผtke (@tobi), Shopify CEO, at Vercel Ship SF: "we are now all verification engineers. We are building harnesses, we are building acceptance tests, and we are buildingโ€ฆ verifiables." When agents write the code, the work moves to the checks: โ†’ harnesses that make agent output repeatable โ†’ acceptance tests that define "done" โ†’ anything a machine can verify without a dev squinting at a diff Full fireside with @rauchg in the reply ๐Ÿ‘‡
3
5
384
S A ๐Ÿ‘‘ retweeted
Vancouver in the fall. The colours. The mountains. The ocean. The towers.
46
217
18
2,320
82,502
S A ๐Ÿ‘‘ retweeted
As long as you are building for humans, you don't need half the speed that the benches are measuring. IO bottleneck will still surface for connected apps. Not all data needs a reactive presentation, that is why virtualization, feedback components like spinners and progress bars exists. The biggest risk in building something humanly useful is in spending too much time optimizing for the wrong things.
Big Rust doesnt want to tell you this but GPUI is in fact not the fastest GUI Library in Rust, not even close.
2
1
5
570
Zero-ar solves 2, still in research preview finesample.com/resources/pdfโ€ฆ Voe solves 4, docs.runvoe.com available for self hosting /Onprem
RFS (request for startups) from a stealth portfolio founder.. I really like all of these ideas. If you are working on any of these, please DM me!
2
42
Stop buying SEO by the task list. Buy it by the outcome. Before you pay any monthly retainer, agree on these four things with the agency, in writing:
1
1
2
130
S A ๐Ÿ‘‘ retweeted
The original plan was to open source my solar installation and share every part of it. Then I asked myself: how useful would that really be to most people? So I did one better. I built a planner that matches your load and budget to a solar design, with the components, where to buy them, and the why behind every decision. And yes, my own system is open sourced too :) I blogged about it here: blog.openculture.org.ng/a-soโ€ฆ The system is live project.anor.openculture.orgโ€ฆ
Some minutes before 9 to 5 starts..but before I switch into work mode.. I just finished something I have been working on for years.. A complete 100% open source of my solar setup. Including detailed schematics, component I used down to every MCB and alternative you can use. where I got them, official distributors I used. I worked with claude over the weekend to push this over the line. I added a complete list of bill of materials. I made it very easy and simple to follow. as you know.. I would be releasing this out with a blog post. Also open source is how I built this so anyone can also use same format to publish their own system. I started this journey 10 years ago knowing zero about electricity.. everything I learned as been someone sharing their knowledge.. this is just me giving back. see you at past 17:01 before then a sneak peak..
32
198
13
575
16,580
"To build AI is to harness cognition itself, and anything so powerful can be used for ill as well as good" ----
39
S A ๐Ÿ‘‘ retweeted
My blog crashed every time it hit the front page of Hacker News. That was my itch, and it's how WP Engine started. All three of my companies started in a similar way -- having a problem myself, being interested in the solution, wondering whether anyone else might agree. So I called other people who ran WordPress and asked what they did when a big wave of traffic took their site down. The answer was always the same: "I don't know. But if you find something, tell me, because I need that." That's interest. It still isn't a business yet. A business requires a lot of other things: Would they pay enough for it? Would they switch now? Do they have the budget already allocated? Is this among their top 3 priorities right now? Can I get to people like this efficiently enough? What langauge do they use to describe this? Where do they already go to find stuff like this? What would I need to show/say to be trusted? What integrations are necessary? So I also asked questions like: "Would you pay $50 a month for it?" I got 30 people to say "yes" before there was anything to sell. Those 30 people weren't like me. I'm the guy who likes setting up a VM, nginx, Varnish and MySQL backups. They'd pay me exactly because they didn't want to do any of that. Your itch is the grain of sand tickling the oyster. It's not the pearl. Go find out if someone will pay you to solve the problem. longform.asmartbear.com/custโ€ฆ
13
3
2
24
2,479
Couldn't have said it better.
You must own your evals to own your alpha. What makes your organization unique can't be protected by benchmarks built for someone else. Ontology Scenarios and Snapshots in Palantir AIP anchor your agents to your enterprise data, letting you freeze any real-world moment, replay it, or stage a what-if to measure performance. Evals then give every agent its own benchmark, on any model.
1
83
S A ๐Ÿ‘‘ retweeted
"As soon as I found out it takes me 300 NOs to get 30 YESes, I no longer was bothered by the nos. My job is now to get more nos than you."
6
20
2
97
2,936
How to see what other people miss. Learn to identify recurring patterns in people, markets, business, opportunities and your own decisions.
15
103
4,112
S A ๐Ÿ‘‘ retweeted
Canadaโ€™s Real-Time Rail (RTR) is liveโ€”a defining step toward a more competitive, resilient and inclusive economy. Read more: payments.ca/canada-launches-โ€ฆ #PaymentsMatter #RealTimeRail
25
65
22
266
27,728
S A ๐Ÿ‘‘ retweeted
The strange reward for admitting you don't know is that you finally get to find out.
30
7
1
96
10,645
A business may not need a general language model for its daily operations once it has clearly written down what its critical parts mean. That sounds far-fetched today. But we think businesses will come to rely on a new class of model, native to the business and to nothing else. General models for the open world, native models for your own business. finesample.com/resources/basโ€ฆ
1
2
73
S A ๐Ÿ‘‘ retweeted
I've always believed that intelligence should come from simple, general building blocks that learn their own abstractions directly from raw data; in other words, the bitter lesson. I think we're finally seeing a real step towards AI that can learn from all data. Language modeling on raw bytes is a microcosm of this ideal, and to me it's a central problem in AI: if we can't build LLMs that learn end-to-end from bytes, how will we handle the rest of the world's data that we understand even less? Last year we introduced hierarchical networks (H-Nets) as a first step. Our new version finds up to 3-5x efficiency gains in controlled scaling law experiments, and at larger scale matches an open-source model trained on 11T tokens with ~10x less data and compute. The posts below walk through what we found and why we think it works.
261
826
288
7,382
1,047,650
S A ๐Ÿ‘‘ retweeted
I'm so excited to finally announce my labour of love for the past few years. Zola and Friends is an animated series about Zola, a white rhino having fun adventures with friends in the Serengeti and Kenya. A few lessons are learned along with the fun. More info out soon.
8
42
2
114
2,952
The ambitious goal is to get an expert/organisation to author meaning (describe all the operations and logic or supply chain for example), and a native model reads with and provides answers, decisions and explanations grounded in how the organization works. So far, solved the OOV problem, today we attempt to resolve the ranking and abstention problem. A model should be able to tell when it doesn't have the right answer instead of inventing one.
39
S A ๐Ÿ‘‘ retweeted
Dark Lisa ๐Ÿ’€๐ŸŽจ
152
1,602
77
15,967
1,234,222